Background: Viscosity is crucial for subcutaneous administration of therapeutic antibodies. In this work, a ProtT5-XL-UniRef50 (ProtT5) and Random Forests (RF)-based prediction method was proposed for accurate viscosity prediction using only corresponding sequences.

Conclusion: GenScript prediction model performs more accurate with higher indicator values on the Ab40 data set.
ACCGenScript = 100%, ACCUM = 70%
mAbs | Experimental Viscosity (cP) | GenScript Predicted | UM Predicted * |
---|---|---|---|
Adalimumab | 12.8 | 12.48 (√ ) | √ |
Atezolizuman | 22.3 | 21.51 (√ ) | X |
Basiliximab | 8.6 | 9.580 (√ ) | √ |
Bevacizumab | 6.7 | 12.73 (√ ) | √ |
Cetuximab | 55.7 | 65.13 (√ ) | X |
Golimumab | 7.6 | 9.760 (√ ) | √ |
Iplimumab | 18 | 20.06 (√ ) | X |
Omalizumab | 60 | 160.6 (√ ) | √ |
Trastuzumab | 9.3 | 10.43 (√ ) | √ |
Ganitumab | 10.9 | 11.17 (√ ) | √ |
* GenScript takes 30cP as threshold and the Exp C are marked according to 30cP
threshold
UM predictions take 15cP as threshold, which means the viscosity value smaller than 15cP are regraded as
LOW (0 as marked in UM Predicted)
Conclusion: GenScript prediction model performs better on the 10 testing samples between GenScript & UM.
The published paper: Hao X, Fan L. ProtT5 and random forests-based
viscosity prediction method …[J]. European Journal of Pharmaceutical Sciences, 2024, 194: 106705.
The comparison method: Makowski E K, Chen H T, Wang T, et al. Reduction of monoclonal
antibody viscosity using … Mabs. Taylor & Francis, 2024, 16(1): 2303781.